山西大学学报（自然科学版）2025，Vol.48Issue(5)：921-932,12.DOI:10.13451/j.sxu.ns.2024030
跨视图一致性表示的多视图属性图聚类
Cross-view Consistent Representation for Multi-view Attribute Graph Clustering
摘要
Abstract
Aiming at the two problems of view embedding representations in GCN-based multi-view clustering algorithms:insuffi-cient cross-view feature consistency and insufficient cross-view clustering consistency,this paper proposes a multi-view attribute graph clustering algorithm(CCRAGC)with cross-view consistent representations.The algorithm strengthens the feature-level con-sistency between view-embedded representations by calculating the node similarity matrix between views,and then constraining the node similarity matrix to approximate the unit matrix;at the same time,it maps the view-embedded representations to the clustering-level subspaces,so that the soft-label matrices in the subspaces are as similar as possible to a way of enhancing the learning of view-embedded representations and correlation of clustering tasks.The results of the study show that CCRAGC is effective for three wide-ly used datasets,namely ACM,DBLP,and IMDB,and Acc improves by 1.21%,0.37%,and 5.74%,respectively,with respect to the benchmark algorithm with the best performance.关键词
聚类/特征一致性/簇级一致性/属性图Key words
clustering/feature consistency/cluster consistency/attribute graph分类
信息技术与安全科学引用本文复制引用
陈晓惠,荆雪纯,曹付元..跨视图一致性表示的多视图属性图聚类[J].山西大学学报（自然科学版）,2025,48(5):921-932,12.基金项目
国家自然科学基金(62376145) （62376145）
山西省科技创新人才团队(202204051002016) （202204051002016）
山西省基础研究计划(202203021212416) （202203021212416）
